Post on 02-Feb-2018
NUCLEAR QUADRUPOLE RESONANCE
FOR EXPLOSIVE DETECTION
Hideo Itozaki and Go Ota
Graduate School of Engineering Science, Osaka University
1-3Machikaneyama Toyonaka, Osaka, 560-8531, Japan
ABSTRACT
Abstract- A Nuclear quadrupole resonance detector has been developed. This detector targets to RDX, a
high explosive inside an anti-personnel landmine, buried up to 15 cm deep. This detector works well
outside an electromagnetically shielded room. It was also mounted an anti-mine vehicle and remotely
controlled mine detection is demonstrated in public.
Index terms: nuclear quadrupole resonance, NQR, RDX, detector, landmine detection.
I. INTRODUCTION
Nuclear quadrupole resonance (NQR) is one of a distinguished candidate for a landmine
detection technique. A widely used metal detector is suffered from high false alarm rate because it
is required to detect just 10 grams of metallic object in a landmine. Highly sensitive metal
detector as this gives an alarm every time it encounters few grams of metal trash, which results in
a bad performance. On the contrary, An NQR detector identifies an explosive inside a landmine
by a resonant frequency unique to each material. The technique will help to take only a bulk of
explosive from metal fragments.
NQR landmine detector has been developed all over the world [1-3] to detect 14N spins of an
INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2008
705
explosive material in a landmine, but very limited paper have been published about the NQR
remote detection of explosive in a landmine. This work reports the development of a prototype
NQR mine detector and test result of its detectability.
II. EXPLOSIVE DETECTION BY NQR
NQR is a kind of interaction between radio frequency (RF) wave and nuclear spins. A
schematic view of NQR detection is shown in Fig. 1.
NQR signal
Excitation RF wave
Equilibrium state
Excitation state
14N
Fig. 1 Schematic view of NQR detection. 14N spin inside a landmine
explosive is excited by RF wave, and then emits NQR signal.
When RF wave with a specific frequency is irradiated, the wave is adsorbed by the nuclear
spins and then re-emitted after the irradiation. Equation (1) shows the NQR Hamiltonian of 14N,
which is the resonant spin in NQR landmine detection [4].
)])((()3([4
222yxyyxxzzzQ IIVVIVeQH −−+−= I (1)
Q is the nuclear quadrupole coupling constant of the resonant spin. , , and are the
spin operator and , , and V are the electric field gradients around the spin to each
directions. Since the electric field gradient is unique to each molecule structure, NQR frequency
is also unique to each molecule.
xI yI zI
zz zz zzV V
HIDEO ITOZAKI AND GO OTA, NUCLEAR QUADRUPOLE RESONANCE FOR EXPLOSIVE DETECTION
706
Fig. 2 shows the NQR frequencies of major explosives, RDX (cyclo–trimethylenetrinitramine),
HMX(cyclo-tetramethylenetetranitramine), and TNT (trinitrotoluene). The difference of the
frequencies allows identifying explosive material. Fig. 3 shows one of NQR signals from 300 g
of RDX (fig.3 (a)) and 300 g of TNT (fig. 3 (b)). Both are detected in an electrically shielded
room. 1,000 data were averaged to improve the signal to noise ratio. Measurement pulse
sequences are the strong off-resonance comb (SORC) for RDX and the spin locking spin echo
(SLSE) for TNT, respectively.
Frequency[MHz]
RDX
HMX
TNT
0 1 2 3 4 5 6
Fig. 2 NQR frequencies of major explosives
(a) (b)
Fig. 3 NQR signals of 300 g of explosives. 1,000 signals were averaged. The signals
were measured in an electromagnetically shielded room.(a) RDX, and (b) TNT
800 8500
10
20
Frequency[kHz]
NQ
R si
gnal
[a.u
.]
3.36 3.38 3.4 3.42 3.440
20
40
60
80
Frequency [MHz]
NQ
R s
igna
l [a.
u.]
The measurement times were 2 second for RDX and 200 second for TNT. This is due to the
difference of the relaxation time of each material. There are two types of relaxation time in NQR.
The one is the longitude relaxation, by which the excited spins are relaxed to the equilibrium state
with the spin-lattice relaxation. This time constant of the spin-lattice relaxation is written as T1.
INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2008
707
T1 dominates the interval time between the sequences, i.e., long T1 drastically decreases the
efficiency of NQR measurement. The other is the transverse relaxation, by which the coherently
excited spins are randomized by the spin-spin interaction and the thermal scattering, resulting in
the cancellation of the NQR signal from each spin. The time constant of the transverse relaxation
is written as T2*. Generally T2
* is shorter than T1. Since NQR signal continues only for T2*, short
T2* makes it difficult to detect the signal.
Fig. 4 shows the relaxation times of RDX, TNT, HMT (hexamethylenetetramine), and PNT
(para- nitrotoluene). HMT and PNT are the raw materials of RDX and TNT, respectively.
RDX
10-3 10-2 10-1 100 1010
1
2
3
T1 [s]
PNT
TNTHMT
T 2* [1
0-3 s
]
Fig. 4 NQR relaxation times of explosives and their raw materials. As shown in fig. 4, RDX has relatively long T2
* and short T1, while TNT has extremely short
T2* and long T1. So RDX is easier to detect than TNT. NQR detector for RDX has been firstly
developed to evaluate NQR detectability in the field.
III. NQR MINE DETECTION
A prototype of an NQR mine detector has been developed by the support from the Japan
Science and Technology Agency. The outlook of the developed sensor-head for NQR mine
detection is shown in fig. 5. The detector, W570mm×D285 mm×H290mm, consists of a sensor
HIDEO ITOZAKI AND GO OTA, NUCLEAR QUADRUPOLE RESONANCE FOR EXPLOSIVE DETECTION
708
coil, a matching box, and some small electrical circuits such as a pre-amplifier. This sensor-head
weighs 10kg, which is prepared to be used by a mine vehicle.
Fig. 5 Prototype NQR mine detector developed with the support from the Japan
Science and Technology Agency.
The performance of the developed NQR detector was evaluated. The NQR signal from RDX
buried in soil was measured. The sample, 100 g of RDX was packed in a cylindrical plastic
case with 110 mm of radius and 80 mm of height. The distance from the bottom of the NQR
detector to the top of the sample case was set to 7, 12, and 17cm. The soil moisture was
controlled 10%. 220,000 data were averaged at maximum to evaluate the relationship of the
measurement time and the signal to noise ratio (SNR). The developed system is capable to
acquire 17,000 data in a minute. The measurement was repeated 7 times every one experimental
condition. One of the NQR signal obtained by 7 cm detection and background noise were shown
in fig.6.
The NQR signal was clearly detected in fig. 6(a). Since the NQR signal certainly appears at a
steady frequency, the signal intensity was evaluated by the output at that frequency. The
dependence of NQR signal intensity on the sample depth was then measured and evaluated. The
environmental noise data were acquired seven times and the averaged background noise height
was calculated in frequency domain. The difference of obtained data and the averaged noise
height was normalized by the averaged noise height. The detection result is shown in fig. 7. The
INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2008
709
square dot of each detection depth shows the average of the seven data, while the error bar shows
the maximum and the minimum of the signal intensity. When the square dot is placed higher than
zero, the signals were detected significantly. If the minimum is placed higher than zero, the
detections were successful enough.
Fig. 6 NQR detection of 100 g of RDX buried 7cm deep in soil with 10% of soil moisture.
Signal accumulation time was 13 minutes. (a) The sample existed.(b) Nothing existed.
3.3 3.4 3.5
0
5
10
Frequency [MHz]
NQ
R s
igna
l [a.
u.]
3.36 3.38 3.4 3.42 3.44 3.46
0
5
10
Frequency [MHz]
NQ
R s
igna
l [a.
u.](a) (b)
Fig. 7 NQR detection of 100 g of RDX buried in soil. The result is normalized by the
averaged noise height. The square dot of each detection depth shows the average of the
seven data, while the error bar shows the maximum and the minimum of the signal
intensity. Dashed line shows the averaged. Signal accumulation time was 13 minutes.
-2
0
2
4
6
8
10
0 10 20
Distance [cm]
(NQ
R s
ignal
-nois
e)/
nois
e
As shown in fig. 7, NQR detection from 7 to 17cm deep was apparently successful. Especially
the sample buried 7 or 12 cm deep seems to be clearly detected. 20 cm detection was, however,
HIDEO ITOZAKI AND GO OTA, NUCLEAR QUADRUPOLE RESONANCE FOR EXPLOSIVE DETECTION
710
almost similar as random determination. This drastic decrease of sensitivity may be due to both
of the signal diffusion and the decrease of the excitation field by the increase of the detection
depth. The latter may be improved by the arrangement of an antenna design and a system
innovation, which will result in the sensitivity improvement of 17 or 22 cm detection.
The result was also analyzed by the receiver operating characteristics (ROC) curve. The
possibility of detection and the false alarm rate were evaluated by the noise levels and the signal
intensities of seven measurements. The ROC curve of 7, 12, 17 cm detection are shown in fig. 8.
22 cm detection was not shown because the data was less outstanding.
0
0.5
1
0 0.5 1
FAR
PD
7 cm
12 cm
17 cm
Fig. 8 ROC curves of the NQR remote detection of 100 g of RDX with the detection
time of 13 minutes.
It is shown in fig. 8 that the sample buried 7 or 12 cm deep were perfectly detected. As few
results have been ever reported about the NQR detection of 100 g of RDX, this result was a
milestone for the realization of the NQR mine detection.
The ROC curve of 17 cm detection was worse than that of 7 or 12 cm detection. This distance
seems to be the limit of current NQR remote detection of 100 g of RDX.
NQR signals detected in 2 minute are shown in fig. 10. The sample was 100 g of RDX buried 5
cm deep (fig. 9(a)) and 10 cm deep (fig.9(b)).
INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2008
711
3.39 3.4 3.41 3.42 3.430
2
4
6
8
10
Frequency [MHz]
NQ
R s
igna
l [a.
u.]
Fig. 9 NQR signal from 100 g of RDX detected in 2 minute. (a) buried 5 cm deep,
(b) buried 10 cm deep. The resonant frequency of RDX is shown by a dashed line.
(a) (b)
3.39 3.4 3.41 3.42 3.430
2
4
6
8
10
Frequency [MHz]
NQ
R s
igna
l [a.
u.]
Though it is easy to see the NQR signal in fig. 9, not so outstanding signal was obtained in fig.
9 (b). 100g of RDX buried 15 cm deep in soil was almost impossible to detect in this time range.
This result shows that the reduction of measurement time is strongly required for the NQR
detection of a deeply buried landmine. Sensitivity improvement will be helpful for the
achievement of this requirement.
IV. DEMOSTRATION OF NQR MINE DETECTION
The NQR detector was mounted on a mine vehicle developed by Hirose Lab. in Tokyo Institute
of Technology [5]. This vehicle can be remotely controlled for the safety of a deminer. Fig. 10
shows the outlook. The detectability was demonstrated in public in September and December ’07.
The integrated mine detection system was so stable that it detected the NQR signal from 100 g of
RDX in 1 minute during the demonstration period.
HIDEO ITOZAKI AND GO OTA, NUCLEAR QUADRUPOLE RESONANCE FOR EXPLOSIVE DETECTION
712
Fig. 10 NQR detector mounted on a mine vehicle developed by Hirose Lab. in Tokyo
Institute of Technology [5].
V. APPLICATION OD NQR DETECTIOR
The developed NQR mine detector is applicable to several fields thanks to its material
detectability. One of expected applications is a security checker. The requirement of security in an
airport and a seaport has increased year by year as the response to the expansion of worldwide
terrorism. The application of NQR detection technology to a luggage inspector has been recently
studied. Prototypes in research are shown in Fig. 11.
Fig. 11 Prototype of NQR luggage checker developed by the cooperation with
Thamway Co., Ltd. in Japan.
NQR detector identifies explosives, which will prevent a terrorist from taking a bomb in an
INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2008
713
airplane or a ship. Not only explosives, but also several narcotics also expect to be detected by
NQR. The realization of these checkers will contribute the social security.
VI. CONCLUSION
An NQR mine detector was reported. The detector detected as small as 100 g of RDX buried up
to 12cm perfectly. The performance evaluation by the ROC curve shows that 17 cm is the
limitation of the detection depth for 100 g of RDX.
The NQR landmine detection was demonstrated in public, in which the detector worked very
well. These results clearly show the possibility of NQR mine detection. The reduction of the
measurement time and extension of detection depth will remain to be realized.
VII. ACKNOWLEDGEMENT
This research was partly supported by Japan Science and Technology Agency.
We thank Prof. Hirose and Dr. Fukushima for the development of a mine vehicle.
REFERENCES
1. R. A. Marino, “Detection and identification of explosives by nitrogen-14 NQR,” Proc. New
Concepts Sym. Workshop on Detection and Identification of Explosives, Quantico, VA, pp. 399,
1978.
2. V. S. Grechishkin and N. Y. Sinyavskii, “New technologies: nuclear quadrupole resonance as
an explosive and narcotic detection technique,” Phys. Usp., vol. 40, pp. 393-406, 1997.
HIDEO ITOZAKI AND GO OTA, NUCLEAR QUADRUPOLE RESONANCE FOR EXPLOSIVE DETECTION
714
3. A. N. Garroway, M. L. Buess, J. B. Miller, B. H. Suits, A. D. Hibbs, G. A. Barrall, R. Matthews,
and L. J. Burnett, “Remote sensing by nuclear quadrupole resonance,” IEEE Trans. Geosci.
Remote Sens., vol. 39, pp. 1108-1118, January 2001.
4. J. A. S. Smith, “Nuclear quadrupole resonance spectroscopy,” J. Chem. Edu., vol. 48, pp.
39-49, 1971.
5. M. Freese, E. F. Fukushima, S. Hirose and W. Singhose,” Endpoint vibration control of a
mobile mine-detecting robotic manipulator,” Proc. 2007 American Control Conf., pp. 6-12, July
2007.
INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2008
715